The Baltimore Longitudinal Study of Aging (BLSA) was established in 1958 and is one the oldest prospective studies of aging in the USA and the world. The mission of the BLSA is to learn what happens to people as they get old and how to sort out changes due to aging and from those due to disease or other causes. Most of our analyses examine the trajectories of performance levels over time. However, we have suspected that there is important information in the inconsistencies in performance, and that greater within individual variability might be associated with subsequent cognitive impairment and dementia. In a case-control sample of Baltimore Longitudinal Study of Aging (mean SD age = 69.90 8.92) participants, we matched 135 clinically diagnosed demented participants with 135 non-demented participants based on age at initial testing and sex. Cognitive performance was examined using measures of memory, executive function, attention, language, and global mental status performance from initial assessment to 5 years before cognitive impairment (mean assessments = 3.03). Compared with unimpaired individuals, individuals diagnosed with dementia had greater variability on measures of attention, executive function, language, and semantic memory at least 5 years before the estimated onset of cognitive impairment, which may indicate maladaptive cognitive functioning. The dementia cases, however, had less variability on visual memory than the unimpaired group, which may suggest that these cases had more difficulty learning. These results suggest that performance variability indexed over annual or biennial visits may be useful in identifying early signs of subsequent cognitive impairment.